In our previous work we found that heritability estimates from related individuals followed a pattern consistent with biases due to shared environment11. In this work we found that a linear additive model, implicitly including both rare and common variants, typically explained less phenotypic variation than that predicted in family studies. These new estimates of narrow-sense heritability are less susceptible to bias and provide additional evidence that family based estimates are inflated. Unlike [11], we were able to obtain estimates for both quantitative and case-control traits. We also found that chip based additive models explained less phenotypic variation than our estimates. In the meta-analyzed phenotypes common to CARe and WHI the average of these estimates were 24.7% and 31.1% respectively. Rare variants and poorly tagged common variants are the most likely explanation for the difference between these two estimates. We discuss other possible explanations below.
